Could someone prove that in a commutative ring $R$ the following two definitions of prime ideal are equivalent:

1) An ideal $P$ is prime if $P\neq R$ and if for all ideals $A,B$ of $R$ with $AB\subseteq P$ we have that $A \subseteq P$ or $B \subseteq P$.

2) An ideal $P$ is prime if $P\neq R$ and if $ab \in P \implies a \in P$ or $b\in P$.

I have read (in Artin as well as online) that these are equivalent, but I cannot find a suitable proof.

Artin's proof of $1) \implies 2)$ is one line: Take $A=(a)$ and $B=(b)$. I do not see how this is a proof. If someone could explain the implication as if I were a complete novice to ring theory (which I am), that would be most helpful.

Note: I am not assuming that $R$ is unital, only that it is commutative.

  • 1
    $\begingroup$ In (2), you would also need to include the condition $P\neq R$, wouldn't you? $\endgroup$ – Arturo Magidin Oct 17 '11 at 1:44
  • $\begingroup$ Which Artin? The one I have (Algebra, Prentice-Hall 1991) explicitly defines a ring to be commutative and unital. $\endgroup$ – hmakholm left over Monica Oct 17 '11 at 1:56
  • $\begingroup$ I am using the same version of Artin. My current professor does not require a ring to be unital, but I still want to use this result. Sorry I phrased it poorly. I meant that the result is true even if the ring is not unital, but I cannot find a clear proof where the author does not assume the ring is unital. $\endgroup$ – nullUser Oct 17 '11 at 2:12

(2)$\implies$(1) always holds: assume that $AB\subseteq P$, and $A$ is not contained in $P$. Then there exists $a\in A$ such that $a\notin P$. But for every $b\in B$, $ab\in AB\subseteq P$, hence by (2) we must have $b\in P$. Thus, $B\subseteq P$. Hence, (2) holds. Note that this holds without the need to assume commutativity of existence of a unity. Condition (2) is called "totally prime" or "completely prime", and is stronger than primality in noncommutative rings.

The problem with (1)$\implies$(2) is just the question of your definition of $(a)$. For commutative rings with unity, $(a)=Ra=\{ra\mid r\in R\}$. However, for commutative rings without unity, since $(a)$ is the smallest ideal that contains $a$, then it is equal to $$(a) = \{ra + na\mid r\in R,n\in\mathbb{Z}\},$$ where $na$ is defined inductively: $0a=0_R$, $(n+1)a = (na)+a$, and $(-n)a = -(na)$. In particular, $a\in (a)$ always holds.

(1)$\implies$(2). Assume that $ab\in P$. Then $(ab)\subseteq P$, hence $(a)(b)=(ab)\subseteq P$. Since $P$ satisfies (1), then either $(a)\subseteq P$ or $(b)\subseteq P$. If $(a)\subseteq P$, then $a\in (a)\subseteq P$ so $a\in P$. If $(b)\subseteq P$, then $b\in(b)\subseteq P$, so $b\in P$. This proves (2).

But perhaps you were thinking that $(a)=Ra$ and so were confused about why this argument would work, given that if $R$ has no unity there is no reason to expect that $a\in Ra$? (it might be there anyway even if $R$ has no unity; e.g., in the direct sum of infinitely many copies of $\mathbb{Z}$ there is no unity, but for every $a$ there is an $x$ such that $ax=a$). The argument as is doesn't work, but the implication still holds: just consider $A$ to be the smallest ideal of $R$ that contains $a$, $B$ to be the smallest ideal of $R$ that contains $b$, notice that $ab\in AB$, and then prove, as I do below, that $AB$ is in fact the smallest ideal that contains $ab$.

The condition that $(a)(b)=(ab)$ is established by noting that every element of $(a)(b)$ is a sum of elements of the form $xy$, with $x\in (a)$ and $y\in (b)$. Each of this is of the form $$(ra+na)(sb+mb) = (rs)ab + (na)sb + ra(mb) + (na)(mb) = (rs + ns + mr)ab + (nm)ab\in (ab).$$ So $(a)(b)\subseteq (ab)$. Since $ab\in (a)(b)$, then $(ab)\subseteq (a)(b)$, which gives the equality.

For not-necessarily-commutative rings, an ideal satisfying condition (2) is known, as I mentioned, as "completely prime" or "totally prime". Completely prime always implies prime, but the converse does not necessarily hold. For an example, take the ring of $n\times n$ matrices with coefficients in a field $F$, $n\gt 1$; the only ideals of this ring are $(\mathbf{0})$ and $R$, so in particular $(\mathbf{0})$ is prime. But it is not completely prime, since for example the matrix $E_{1n}$ ($1$ in the $(1,n)$ coordinate, zeros elsewhere) satisfies $E_{1n}^2 = \mathbf{0}\in (\mathbf{0})$, but $E_{1n}\notin(\mathbf{0})$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.